The 'real' Premier League table decided by xG

[The Doctor]

38 minutes ago, Sampson said:

Be interesting to see the xG for GF vs GA, my instinct is it’s the other way round and we conceded too many soft goals to poor quality chances 

depends on the model, but I use understats data: https://understat.com/league/EPL

 

xG differential was -2.01 (i.e. our xG was 2 lower than our actual GF), xGA differential was +8.21 (i.e. conceded 8 more than expected)

[Sampson]

Just now, The Doctor said:

depends on the model, but I use understats data: https://understat.com/league/EPL

 

xG differential was -2.01 (i.e. our xG was 2 lower than our actual GF), xGA differential was +8.21 (i.e. conceded 8 more than expected)

Yeah that’s what I expected.  Didn’t realise we actually scored more than our quality of chances suggests, but we concede way more than an average defence vs those chances.

[Number 6]

1 hour ago, Death by Football said:

So does actual goals scored.

That is by definition, xG predicts future performance better than actual goals though.

[Number 6]

57 minutes ago, HitchinFox said:

Yep, Brentford are brilliant at using data to drive everything. And they use xG in the right way – to discover individuals who can finish and to identify players with "hidden value". And yes, I'm sure xG has played a part in their success. Equally though, their entire "Moneyball" approach is much more than just xG. They do a lot of other data-based research and analysis outside of it. 

My point is that – and again, just my opinion – xG has become overused these days, and is often applied in ways that don't really work. Especially when it comes to predicting team performances as a collective over a long period of time. For example, speculating about where teams should have "really finished" in a 38-game league based on xG. That's not really what xG is for. 

I'm all for data – in fact, a bit of a geek when it comes to it. Just feel that xG has become this thing that is used for everything. 

Maybe that's more it, actually. I just see it being everywhere. It is useful, 100%. But over-used and often in the wrong context. 

Completely agree, and I think this is the bit that annoys people, the overuse and the phrase "expected". Doesn't help that teams like Brentford create their own "Justice League" which implies actual performance is less relevant. Should be called weighted shots or something I think.

 

Looking at xG data really did show the drop off we had under Brendan though, and why underperformance this season shouldn't have been a huge surprise.

[LaCiudad]

Shows what crap xG is

[Duncan_K]

One thing to remember with this is we were 14th on the actual goal difference table too. So it's not just about we should have scored more or conceded less, it's about bottling important games and not doing enough to actually get points on the board, essentially far too many 1 goal defeats. One of the saddest things in all of this is just how many teams with appalling goal difference we've let stay up ahead of us. 

[Les-TA-Jon]

1 hour ago, Death by Football said:

Such utter codswhallop!

The fact that in almost every game the xG is different to the actual goals scored shows that xG is usually wrong.  Therefore it's a really useless metric that distracts from ineffective performances.  I could have a massive xG in a game because I keep going through one on one with the keeper but my goals scored will be nil because I'm fairly sh!t at football.

 

1 hour ago, Death by Football said:

So does actual goals scored.

 

13 minutes ago, LaCiudad said:

Shows what crap xG is

Hilarious watching all the dinosaurs moaning about xG in this thread. It's a stat. It's not meant to be a silver bullet. It's just another tool in the box for analysing the game. 

 

Do people over use it or put more stock into than they should? Yes. 

Is it a totally pointless stat that has no value and we should just use 'actual goals scored' instead? No.

 

It's just a way of quantifying the number and quality of chances a team had. 

 

You see a score line like Leicester 0, Trumpton Town 4 and think Leicester got battered. 

 

You see a score line like Leicester 0 (xG 2.1), Trumpton Town 4 (xG 2.5) and you see that it was a somewhat anomalous result and have fuller picture of the scoreline (i.e it's quantified, not just a "I reckon") without even having to see the match or highlights

[Number 6]

8 minutes ago, Les-TA-Jon said:

Hilarious watching all the dinosaurs moaning about xG in this thread. It's a stat. It's not meant to be a silver bullet. It's just another tool in the box for analysing the game. 

 

Do people over use it or put more stock into than they should? Yes. 

Is it a totally pointless stat that has no value and we should just use 'actual goals scored' instead? No.

 

It's just a way of quantifying the number and quality of chances a team had. 

 

You see a score line like Leicester 0, Trumpton Town 4 and think Leicester got battered. 

 

You see a score line like Leicester 0 (xG 2.1), Trumpton Town 4 (xG 2.5) and you see that it was a somewhat anomalous result and have fuller picture of the scoreline (i.e it's quantified, not just a "I reckon") without even having to see the match or highlights

My favourite example of this is from Brentford 1-3 Man Utd last season, where the xG data was really interesting.

 

Frank came out after the game saying they 'destroyed' Man U in the first half and the result was unfair, which is somewhat fair given the stats were 1.48 v 0.19 on xG in the first half. But he also really downplayed Man U in the second half.

 

xG overall was pretty much 2-2, but this ignores the fact that the two biggest single chances both fell to Utd (0.57, 0.64) and both were scored. They were the only chances in the game above 0.5xG. 

 

Also Brentford's xG was created from 18 shots, to Utd's 13. So xPoints would have favoured Utd as their average shot quality was higher. Brentford's best chances fell to Jensen, who didn't score all season in the league, Utd's fell to Greenwood and Elanga. It was a good away performance from Utd, who maybe did get slightly fortunate to be 0-0 at half time.

 

The stat isn't without its flaws, but you can tell a really compelling story with it sometimes. Pundits and the media really misuse it though.

 

Data for the game here.

[Death by Football]

18 minutes ago, Les-TA-Jon said:

 

 

Hilarious watching all the dinosaurs moaning about xG in this thread. It's a stat. It's not meant to be a silver bullet. It's just another tool in the box for analysing the game. 

 

Do people over use it or put more stock into than they should? Yes. 

Is it a totally pointless stat that has no value and we should just use 'actual goals scored' instead? No.

 

It's just a way of quantifying the number and quality of chances a team had. 

 

You see a score line like Leicester 0, Trumpton Town 4 and think Leicester got battered. 

 

You see a score line like Leicester 0 (xG 2.1), Trumpton Town 4 (xG 2.5) and you see that it was a somewhat anomalous result and have fuller picture of the scoreline (i.e it's quantified, not just a "I reckon") without even having to see the match or highlights

You make a fair point - the overuse of all these statistics is partly what gets me.  it has it's place being used by expert analysts to try and improve a team.  It is meaningless to the majority of fans who are shown it applied to irrelevant outcomes like the table this thread is based on.  My initial response was based on the table, not the use of xG in itself.

One of my dislikes it is that it's a loser's statistic.  When we win a game how many people give a toss what our xG was compared to actual goals scored? Lose a game and it's used to make everyone feel better about themselves as in your example above.  It's not relevant whether we were quantifiably battered or subjectively battered - we still lost 4-0 at home to Trumpton Town. 

[lcfc_forever]

42 minutes ago, LaCiudad said:

Shows what crap xG is

Why do you think it's crap?

 

It validates what we all knew - we conceded more goals than we should have done. And why was that -- we had a crap goalkeeper, and his successor while good, made major errors especially vs Fulham and Liverpoool.

 

I think the xG vs Fulham was something like 4-2 in our favour yet we lost 5-3. 

 

As a lot of our posters have said, we were just too easy to score against until it was too late. 

[Tuna]

Xg is a load of old cobblers.

[Death by Football]

3 minutes ago, lcfc_forever said:

Why do you think it's crap?

 

It validates what we all knew - we conceded more goals than we should have done. And why was that -- we had a crap goalkeeper, and his successor while good, made major errors especially vs Fulham and Liverpoool.

 

I think the xG vs Fulham was something like 4-2 in our favour yet we lost 5-3. 

 

As a lot of our posters have said, we were just too easy to score against until it was too late. 

Kind of sums it up.

[Les-TA-Jon]

3 minutes ago, Death by Football said:

You make a fair point - the overuse of all these statistics is partly what gets me.  it has it's place being used by expert analysts to try and improve a team.  It is meaningless to the majority of fans who are shown it applied to irrelevant outcomes like the table this thread is based on.  My initial response was based on the table, not the use of xG in itself.

One of my dislikes it is that it's a loser's statistic.  When we win a game how many people give a toss what our xG was compared to actual goals scored? Lose a game and it's used to make everyone feel better about themselves as in your example above.  It's not relevant whether we were quantifiably battered or subjectively battered - we still lost 4-0 at home to Trumpton Town. 

I don't disagree that it can help fuel a loser's mentality. 

 

I totally disagree here though - of course it's relevant!? How teams win/lose games is relevant, for a whole host of reasons.

 

Great Escape season as an example: We were 1 of only 2 teams (along with Arsenal who finished 3rd on 75 points) that season to never lose by more than 2 goals. We kept losing, but we were competitive for much of the season. Had we been continually spanked, I'm not sure we'd have had the platform/confidence for the Great Escape itself

[Death by Football]

4 minutes ago, Les-TA-Jon said:

I don't disagree that it can help fuel a loser's mentality. 

 

I totally disagree here though - of course it's relevant!? How teams win/lose games is relevant, for a whole host of reasons.

 

Great Escape season as an example: We were 1 of only 2 teams (along with Arsenal who finished 3rd on 75 points) that season to never lose by more than 2 goals. We kept losing, but we were competitive for much of the season. Had we been continually spanked, I'm not sure we'd have had the platform/confidence for the Great Escape itself

But that example is a totally different point.  It's not relevant to xG it's relevant to the actual results.  You say we were competitive for much of the season because we never lost by more than two goals and that bred confidence.  Not that our xG was consistently similar to or higher than opponents that beat us by large margins.  

[Muzzy_no7]

Lack of consistency at the back in terms of selection due to injuries and Brendan being a kn*b has cost us relegation. 

[Les-TA-Jon]

1 minute ago, Death by Football said:

But that example is a totally different point.  It's not relevant to xG it's relevant to the actual results.  You say we were competitive for much of the season because we never lost by more than two goals and that bred confidence.  Not that our xG was consistently similar to or higher than opponents that beat us by large margins.  

I don't know what our xG and xGA was like that season, but the point is, would you rather all your losses be with 0.0 xG or 2+ xG. I felt like you were basically claiming there's no difference between the two? 

 

Yes they both result in 0 points, but it's the old cliche about losing well.

[Death by Football]

7 minutes ago, Les-TA-Jon said:

I don't know what our xG and xGA was like that season, but the point is, would you rather all your losses be with 0.0 xG or 2+ xG. I felt like you were basically claiming there's no difference between the two? 

 

Yes they both result in 0 points, but it's the old cliche about losing well.

I've forgotten what it's like to lose well.

[lcfc_forever]

If the people in control of the club based their decisions on analytics like xG, they would have got rid of BR earlier seeing how poorly we were performing. A piece in the Athletic suggested our xG had deteriorated around Jan/Feb to 1-2, which is the latest we should have removed BR. 

 

They would have also heavily invested in a top goalkeeper in January looking at the stats. They did neither and look what happened. 

 

This was all so avoidable. 

 

Only possible consolation is they learn from this and don't go into a season with such complacency over a pivotal position in the team. 

Edited 1 June 2023 by lcfc_forever

[Finnegan]

2 hours ago, HitchinFox said:

to see a high xG and then to say that Team X "should've won the game" or "finished higher in the table" is silly, if the shots were taken by, say, Patson Daka. 

 

Patson Daka outperformed his non-penalty xG per 90 minutes over the last two years by more than Kelechi Iheanacho or Jamie Vardy. He's the only of our three strikers to outperform his xG in both of those two seasons. Ergo, contrary to what I'm reading as a slightly snide implication (apologies if I've misunderstood your intent,) Patson Daka is the striker you'd MOST want to be presented with a good chance in a Leicester shirt. 

 

As for the rest of your post, xG isn't and wasn't designed for showing which team "deserved" to win or "should have" won, it's designed to show how productive a team were in terms of creating chances. 

 

You're absolutely correct that it doesn't reflect the ability of the finisher and nor should it, that's exactly why you then want to compare it to other indicators such as that player's np:xG-aG as above. 

 

The problem with stats in football is that they're not really "for" everyone. And I don't mean in the sense that they're not everyone's preference, I mean that not everyone understands what they're looking at. 

 

My actual 9-5 day job is data analytics (albeit in a sector far more mundane than football) and a big part of my job isn't just producing data, it's explaining it to senior decision makers in the business. Even some of the most competent, brilliant people I know aren't great at looking at a bunch of numbers and framing them in context that help them make informed business process decisions. That's why they pay people like me and why people like me are employed by every successful company in the world. 

 

When it comes to football, Bob down the pub thinks he's an expert in football, he think he knows the game, everyone down the Dog and Duck thinks he's the local footy expert. So Bob just assumes he should understand the stats. But he doesn't because stats and data aren't his thing because they take a reasonable amount of expertise. Bob's ego can't hack this so he hand waves stats off and just starts waffling on about how xG is a load of rubbish and how football is played on the pitch instead of on a spreadsheet. 

 

Thing is, it isn't just Bob down the pub. It's everyone from Graeme Sounnes to Michael Owen to Jamie Carragher to Rio Ferdinand. The TV is full of these guys that don't understand data and some of them are dismissing it as a result and others are misusing it, confusing everyone and giving ammunition to those that would belittle it. It's all over social media exactly the same, too. 

 

I dont mean to imply I'm a genius or that it takes a rocket scientist to understand some fundamental football stats but they definitely aren't for everyone and the expectation shouldn't be that just because football is accessible, football data analytics also is. There's a reason every professional club on the planet is employing some seriously well educated, experienced data analysts to study every part of the game. They will put a lot of value in modern metrics like xG, it's just that they aren't using it the same way some spotty knob head wannabe youtuber is when he's trying to convince everyone why Tottenham deserved to win X or Y game because Harry Kane took a bunch of pot shots. 

[The Doctor]

46 minutes ago, Death by Football said:

.

One of my dislikes it is that it's a loser's statistic.  When we win a game how many people give a toss what our xG was compared to actual goals scored? Lose a game and it's used to make everyone feel better about themselves as in your example above.  It's not relevant whether we were quantifiably battered or subjectively battered - we still lost 4-0 at home to Trumpton Town. 

almost like it's intended use is to track longer term performances. unless you are world class (Haaland) or completely out of your depth (Ward), your performances will track back towards the mean. measuring xG over a run of games helps indicate over/underperformance to then advise coaches on issues when broken down into more granular data (by game situation (winning, drawing, losing), specific chance creation routes (e.g. are you drastically underperforming from set pieces?) etc. 

 

people not understanding it doesn't make it a bad statistic

[Finnegan]

15 minutes ago, The Doctor said:

People not understanding it doesn't make it a bad statistic

 

I want Mark to add this to foxestalk's language filter so that every time someone types the letters "xG" it's replaced with "xG (people not understanding it doesn't make it a bad indicator)" 

 

[Death by Football]

16 minutes ago, The Doctor said:

almost like it's intended use is to track longer term performances. unless you are world class (Haaland) or completely out of your depth (Ward), your performances will track back towards the mean. measuring xG over a run of games helps indicate over/underperformance to then advise coaches on issues when broken down into more granular data (by game situation (winning, drawing, losing), specific chance creation routes (e.g. are you drastically underperforming from set pieces?) etc. 

 

people not understanding it doesn't make it a bad statistic

I agree.  It's intended use is for people in the game who understand it, can analyse it and use their findings to advise coaches on how to improve issues.  

I have no issue with xG or any other xStatistic, only with the fact that I am constantly bombarded with them when there is nothing I can do with that data.

[Stadt]

3 hours ago, HitchinFox said:

This is one of the examples of why xG should only ever be used as a rough guide. In my humble opinion. 

There is, rightfully, a lot of data-crunching now in modern elite football. Plenty of things that can be studied, learnt from and applied in training to correct and improve players, tactics etc. But not sure xG is one of them. 

 

Also, adding to the point made by @beepee1984 there are other issues with it.

After all, xG is, at its heart, designed as a metric that's intended to measure the probability of a shot resulting in a goal – it takes into account the place it was taken from, where the pass came, which body part (head or leg) etc. It is pretty clever, no doubt. 

But - and here's the massive but. What it doesn't take into account, at all, is whether that shot was taken by someone like Erling Haaland, Mo Salah or a donkey called Tripod. It purely looks at the quality of the chance. Not the quality of the player (which, to be fair, could be subjective and ruin the data). 

So to see a high xG and then to say that Team X "should've won the game" or "finished higher in the table" is silly, if the shots were taken by, say, Patson Daka. 

In essence, xG only ever tells us what we already know by watching a game – whether a team is converting its chances. Those who passionately believe in xG's mythical powers often say that it is, and I quote: "highlighting teams who are over or under-performing their expected numbers and whose results may soon begin to change."

Well, yes and no. A team might keep producing chances but results are unlikely to change or improve, if the same, not-very-good players are still taking the shots? However, if a star striker has been out injured and his understudy has been faffing chances, then yes – once the main man comes back, results are likely to improve and the actual goals are more likely to match xG?

But we already probably knew that without the data that xG gives us?

I'm prepared to be convinced that xG is the best thing since slide bread and open to changing my mind. But as it stands, I do think it's use is limited and definitely not as powerful an analytic tool as some make out. It's also a source of some very lazy journalism, which is a particular pet-hate of mine. 

Very few players consistently outscore their xG. I can't find the graph now but finishing ability between players is surprisingly similar. This is probably the best article I can find https://statsbomb.com/articles/soccer/quantifying-finishing-skill/

[Finnegan]

7 minutes ago, Death by Football said:

I agree.  It's intended use is for people in the game who understand it, can analyse it and use their findings to advise coaches on how to improve issues.  

I have no issue with xG or any other xStatistic, only with the fact that I am constantly bombarded with them when there is nothing I can do with that data.

 

 

Personally I love it, I don't care what I can and can't "do" with data I just find it fascinating to have. It's of interest to me in the same way, idk, stamp collecting or train spotting are to other people. It's just a hobby. I personally don't understand why people get excited about TIFO or which European ultras did some crazy shit this week, it's just everyone takes something slightly different from the sport and it's cool that it can keep us all entertained in different ways. 

 

Really, though? Bombarded? You don't think that's slight hyperbole? I know xG is very vogue at the moment but I find it kinda easy to avoid nearly all of this sort of stuff. Idk, maybe I'm just not very active on social media outside this forum but I don't have a problem avoiding things I'm not interested in. I don't, for example, bother reading the Ultras thread on foxestalk all that much and I wouldn't make a habit of go in there and rolling my eyes at them all regularly. I think if you choose to come in to a thread that's clearly about xG and then go on to whinge about it when you made the decision to read this knowing full well what it is then, eh. That's on you, no? 

[The Doctor]

14 minutes ago, Death by Football said:

I agree.  It's intended use is for people in the game who understand it, can analyse it and use their findings to advise coaches on how to improve issues.  

I have no issue with xG or any other xStatistic, only with the fact that I am constantly bombarded with them when there is nothing I can do with that data.

to influence the club sure, none of us are employed as analysts for Leicester City Football Club, but it's really useful if you like a flutter for instance. 

 

also, it's very interesting stuff imo, if you don't find it interesting then it's very easy to avoid